US 7,031,538 B2 · Method and apparatus for feature-based quantization and compression of data

Abstract

A method and apparatus for compression and restoration of images, videos, surfaces, curves and, generally, pixel-based data sets, including: feature-based quantization using a set of N isocontours; compressing each of these N contours using a level set-based compression of curves or surfaces—and/or combining the quantization with a conventional method for compression of surfaces; and decompressing the compressed data using a nonoscillatory reconstruction. The method is naturally multiscale, but does not use wavelets or other multiscale-basis functions. The method performs compression and decompression (1) without excessive computational complexity, even in the presence of topological changes such as merging or breaking of isocontours as the scale changes, (2) with excellent image fidelity, (3) without unacceptable artifacts such as ringing, blurring, or blocking, (4) with simple and natural grid based ways of calculating geometric features such as normals, principal curvatures.

Claims5

Dependent claims

A method for quantizing and compressing an individual image, including the steps of:

selecting a set of contour curves for the image;
generating a representation of the image correlated with the contour curves and based upon at least one predetermined feature-oriented criterion of the image;
storing the generated representation of the image;
generating a further representation of the stored image as a zero-level contour;
and
generating a multiresolution representation of the zero-level contour.

3 more elements

Dependencies are inferred from the claim text.

Method and apparatus for feature-based quantization and compression of data

Life of this patent

Method and apparatus for feature-based quantization and compression of data

Life of this patent

Abstract

Claims5